Skip to main content
Dryad logo

Data from: Spatiotemporal dynamics of the ant community in a dry forest differ by vertical strata but not by successional stage

Citation

Neves, Frederico et al. (2020), Data from: Spatiotemporal dynamics of the ant community in a dry forest differ by vertical strata but not by successional stage, Dryad, Dataset, https://doi.org/10.5061/dryad.2bvq83bp8

Abstract

Ants are diverse and ecologically important organisms in tropical forests, where their spatiotemporal distribution can be highly complex. This complexity arises mainly from marked differences in microclimatic conditions and resource availability through space and time that is even more evident in highly seasonal environments, such as tropical dry forests. However, it is unclear how seasonality interacts with other factors that might shape temporal variation of ant composition (β-diversity), like vertical strata and habitat disturbance. Our goal was to examine the potential influence of vertical stratification and the successional stage on the spatiotemporal variation of a tropical dry forest's ant species composition. We assessed whether species turnover or nestedness was the main component determining the spatiotemporal β-diversity of ant communities across the canopy and litter strata. We sampled canopy and litter ants in ten plots, half in the early and half on the late-stage of secondary succession at four times, twice in wet and twice in dry season. A high species turnover defined the spatiotemporal β-diversity of canopy and litter ant communities across years and seasons in our focal dry forests. Importantly, the temporal ant species composition was much more stable in the canopy than in the litter. Moreover, we found that the ant community’s temporal dynamics was consistently high across successional stages, not differing in the temporal β-diversity between early and late succession. Our results provide valuable insights into the potential underlying causes of community assembly and spatiotemporal dynamics in seasonal habitats, like the highly-threatened and diverse tropical dry forests.

Methods

Study area

We conducted this study in Parque Estadual da Mata Seca (PEMS; 14°48′36″ – 14°56′12″ S and 43°55′12″ – 44°04′12″ W; 493 m a.s.l.), in SE Brazil. The PEMS is located in a transition zone between three major Brazilian biomes: Cerrado (Brazilian savanna), Caatinga (spiny-dry forest), and Mata Atlântica (Atlantic Forest). With a total area of 15,466 ha, the PEMS is inserted in a mosaic of primary and secondary (previously used for agriculture and livestock activities) forests (Madeira et al., 2009). The climate of the region is tropical semi-arid (Alvares et al., 2013), characterized by a severe dry season from May through September and a wet season from November through April. The average annual precipitation is 818 ± 242 mm (mean ± SD), which is concentrated in the wet season months, and the average annual temperature is 25.1 °C (Pezzini et al., 2014).

The whole study area was covered originally with tropical dry forests on fertile soils and mostly level terrain. The tropical dry forest of the area is comprised of deciduous vegetation that loses more than 90% of its leaves in the dry season (Pezzini et al., 2014). We chose areas belonging to two successional stages, hereafter early- and late-stages, with remarkable differences in their structure and land-use history (Madeira et al., 2009). The early-stage area was a pasture and has been in recovery for about 15 years. The early-stage contains mixed vegetation of trees less than 8 m in height, herbs, and grasses, without distinct vertical stratification. The late-stage area has had little to no anthropogenic disturbance for at least the last 60 years, with emerging trees that exceed 20 m in height. Moreover, the late-stage area has distinct canopy and understory strata, with direct sunlight reaching the soil level only in the dry season when the trees have lost their leaves. The late-stage areas also have low densities of new trees and lianas compared to other adjacent areas at different stages of regeneration (for a detailed description of the study areas see Madeira et al., 2009).

Ant sampling

We sampled ants in the canopy and litter strata in ten 20 m × 50 m rectangular plots (Figure 1). These plots were spaced at least 200 m apart along a 7 km transect, with five in the early-stage and five in the late-stage of secondary succession. Plots from the same successional stage were located 0.2–1.0 km from each other (Antoniazzi et al., 2019; Madeira et al., 2009). All ten plots were subjected to similar topographic and climatic conditions, regardless of their successional stage (see Madeira et al. 2009). We acknowledge that the spatial distribution of our plots represents a pseudoreplicated design (i.e., ‘clumped segregation’; sensu Hurlbert 1984), but it is not considered a serious concern when monitoring natural experiments where site and disturbance/treatment effects may be confounded (see Davies & Gray, 2015, Oksanen, 2001). In each plot we chose a grouping of three trees at each of the four corners and in the center of each plot to sample ants, giving a total of 15 trees and litter samples per plot. The plots were sampled four times, twice in the dry season (September 2010 and September 2011) and twice in the wet season (February 2011 and February 2012). The ant diversity in each stratum over the four periods was then used to calculate the beta-temporal diversity, as described below.

We used specific methods for sampling the ants in each stratum. In the canopy, we used two complementary techniques: arboreal pitfall trapping and an entomological beating technique (Antoniazzi et al., 2019; Campos et al., 2006). Arboreal pitfall trapping consisted of placing 15 cm diameter plastic pots filled with water and soap and collecting after 48 hr. The entomological beating technique consisted of using an entomological umbrella made from an inverted cloth funnel, with a surface of 1 × 1 m, and a plastic bag attached to the bottom (for more details, see Antoniazzi et al., 2019). To sample the ground-dwelling ants, we used the Mini-Winkler method, a widely used technique to collect arthropods that live in the litter (Belshaw & Bolton, 1993; Castro et al., 2012; Silva & Brandão, 2010). At each sampling point below a tree, the litter from 1m² was collected and sieved through 1cm² mesh. The filtered material was then placed in the Mini-Winkler extractor that was suspended for 48 hours (see Fisher, 1999) so that the associated fauna fell into the collecting pot.

We transported all the sampled material to the laboratory for sorting, and the ants were mounted and identified to the lowest possible taxonomic level. First, we identified all ants to genera using Baccaro et al. (2015). For species identification, we then used a combination of the keys of “Ants of Costa Rica” (Longino, 2007), and the information and images available on “AntWeb” (Fisher, 2002). For the current valid taxonomic name of each species, we consulted “AntCat” (Bolton, 2012). Our species determinations were then further verified by a reference collection from previous work in the same area (Antoniazzi et al., 2019). We deposited all material in the Laboratório de Ecologia de Insetos (LEI), Universidade Federal de Minas Gerais (UFMG), Brazil.

Statistical analysis

Firstly, we assessed the sampling completeness of the ant species for each successional stage separately. For this, we used the rarefaction and extrapolation (the number of samples doubled) of ant species richness (Chao et al., 2014), since the ant fauna was recorded using Winkler in the litter and arboreal pitfalls and beating at the canopy. The sample units (i.e., ten plots per time period and four time periods, totaling 40 samples) were used as the independent variable and ant species richness was the dependent variable. Thus, we were able to obtain information about the completeness of ant species in relation to each sampling unit added. We also assessed the alpha diversity in our smallest sampling unit for each successional stage apart, i.e., early and late. For this we used the hill number 0D to show the observed species richness accumulated for our sampling units, with their confidence intervals. The overlapping of confidence intervals indicates that the two treatments present the same species richness. For the sampling coverage and the species richness, we used the package iNEXT (Hsieh et al., 2016) for R software.

We assessed the spatial and temporal variation of the ant species composition (β-diversity) using the Sørensen dissimilarity in a multiple-space and time approach (see Baselga, 2010). For spatial analyses, the multiple samples from each stratum, i.e., litter and canopy, were pooled per plot for subsequent analyses. For temporal analyses, the multiple samples from a stratum within a plot were pooled for each time period, to give four plot-level data points for each stratum and each plot for subsequent analyses. We then calculated the spatial and temporal β-diversities (βSOR) using the presence/absence of species for each stratum and plot (10 plots in total) for spatial analyses, and over the four sampling periods for each stratum and plot (10 plots in total) for temporal analyses. The β-diversities (spatial and temporal) were further partitioned into the components derived from species turnover (βSIM) and the derived component of nestedness, i.e., species gain/loss (βNES), using the β-diversity partitioning method components. We used the "beta.multi" function of the R package betapart to the partition of spatial and temporal β-diversities (Baselga & Orme, 2012).

Finally, we used the temporal β-diversity (βSOR) to test the predictions that the temporal β-diversity is higher in the litter and early-stage than in the canopy and late-stage. To compare the temporal β-diversity values we used a generalized linear mixed model (GLMM) with stratum, successional stage and the interaction between these variables as the fixed effect variables, and the identity of each successional plot as a random factor to account for spatial pseudoreplication (Bolker et al. 2009). We used the Gaussian error distribution, and we checked for the error distribution and over-dispersion of the data (Crawley 2013). We defined the minimal model after the removal of non-significant variables (P > 0.05). We performed all analyses using the R software environment for statistical computing (R Development Team, 2019).

Funding

Conselho Nacional de Desenvolvimento Científico e Tecnológico

Fundação de Amparo à Pesquisa do Estado de Minas Gerais

National Science Foundation, Award: 1442256